Showing posts with label speech recognition. Show all posts
Showing posts with label speech recognition. Show all posts

Tuesday, May 13, 2008

Recognition for Speech Recognition

My friends came through again. If you've been following this blog you may remember reading about my attempts to get dictation software installed and running. I ended up creating a virtual Virtual PC. Unfortunately, the package with the IBM ViaVoice did not contain the CD. However, yesterday I received a disc in the mail from my friend; ViaVoice version 10 for Windows XP.

This is the first article that I am attempting to dictate. So far the recognition has been so-so. Correcting mistakes is not intuitive yet. I'm sure it will get better as I get used to the software. When it works the speed is amazing, so I am looking forward to getting this working better.

In order to make this somewhat more technical I am going to describe the set up that I am using in the order to get my dictated text onto the Mac.

ViaVoice's SpeakPad dictation application.

As I have it set up, ViaVoice allows dictation into its own application called SpeakPad. I am using this program to create a simple plain text file. In order for my Mac to access it, I set up Windows file sharing. Because I am using the Powerbook as a wireless gateway for the PC, it was somewhat harder to get the two computers to see each other. I was able to see the Windows share from my Mac but not vice versa.

The simple green screen interface of WriteRoom.

Aside from that issue setting up sharing was fairly straightforward. I followed the directions here. Then, to finally published the article I am using several Mac applications. First, I save the file from SpeakPad into the shared directory. I use the application, WriteRoom, to open that text file in order to proofread. I like the green and black full screen mode for this because it is easy on my eyes and typing with text expansion is responsive. Plus, once a file is open in WriteRoom, it will reopen it the next time the program is run. I simply save over the same text file in the virtual PC, and the Mac text editor reflects those changes. Next, I drag the text file into MacJournal in order to keep a nice, searchable copy. From there it's one click to publish. I open the blog in a browser where I add images and links and do the final check.

It seems that voice recognition will be a real time saver if I manage to tame it. Obviously the current setup is somewhat convoluted, but moving the information from app to app is actually fairly painless. And the much smaller amount of typing certainly causes me less pain. It's nice to have good friends, and it's especially good to have nice friends.

Sunday, March 16, 2008

Making a Virtual Virtual PC


Before the transition from PowerPC to Intel CPU's by Apple, Mac users had to rely on various other means if they needed to run the occasional Windows app on their computer. The most popular and practical solution was the commercial software package, Virtual PC. It was created by the company, Connectix, but subsequently purchased by Microsoft. Virtual PC would emulate a full Pentium II computer system. Unfortunately, translating instructions between processors is time-consuming and fraught with errors. The program worked for the most part, but was slow at best.

Recently, some of my friends conspired to get me voice recognition software. As I have posted in the past, here and here, my ability to type has been greatly decreased, and I have sought ways to make entering information into the computer faster and easier. Dictation is certainly a good option. It so happened that my friend had gotten a free copy of IBM's ViaVoice for Windows. The system requirements call for a 600MHz Pentium III. Obviously, my emulated Pentium wouldn't cut it.

The native Mac options for speech recognition are very limited. The company, MacSpeech, makes a product called iListen that would run on my G4 PowerBook. There is also a Mac version of ViaVoice. Unfortunately, these products have gotten fairly negative reviews and cost between $50 and $100. A new, superior product called Dictate is being released by MacSpeech, but it will only run on Intel Macs. The answer to this conundrum, my friends figured, was to just send me another computer.

So I have inherited my friend's Gateway 1.6GHz Pentium 4 with Windows XP. Since he wasn't using the computer, this became the most economical option. However, there were a couple of issues to overcome in order for this system to be useful to me.

Issue one: No wireless
First of all, my computer is set up in the garage, far away from any physical network cabling. I use the Mac's Airport card to connect to our wireless network. The Gateway, being an older desktop model, has no wireless available. Thanks to OS X's amazing networking, though, this problem was easily resolved. The first step was to connect the Gateway and the Powerbook with an ethernet cable. Normally this kind of direct connection would require a crossover cable, but I recalled reading that Mac network adaptors automatically detect when a crossover cable would be needed and adjust accordingly to use a regular cable, so I gave it a try. The next step was to enable Internet sharing in the Mac's System Preferences. This was as simple as choosing Airport from the drop-down list and clicking Ethernet for the other computer's connection. I hooked up the cable, started the sharing, and it just worked.

OS X Tiger Internet Sharing

Issue two: Only one monitor
The second issue was that I only have one monitor available (and really no room for another). Normally, I could just use my laptop's display and hook the monitor to the PC. Unfortunately, due primarily to some very rough treatment by its previous owner, my Powerbook's hinge broke, disconnecting the LCD from the computer internals and effectively rendering the display useless. Therefore, sharing the one monitor was the only option.

For several reasons I decided to continue using the Mac as my main computer and run the Gateway remotely as needed. To facilitate this I installed VNC on both systems. For the Windows VNC server, I used Alkit VNC, which allows the sharing of a single application window or the whole screen. I figured this would enable me to just have a ViaVoice window or IE window displayed within the laptop's applications. On the client side I used JollysFastVNC. Though technically just an Alpha release, this app eliminates the biggest problem in using VNC; this client is actually fast. In my limited use I have not had any problems with it. You can see the Gateway's XP desktop running half size in the screenshot beginning this post.

Final issue: No disc!
After getting everything set so I had a virtual Virtual PC, a real PC in a VNC window, it was time to install ViaVoice, the whole reason for all this effort. When I opened the box, I found a nice USB headset mic and a directional desktop mic, a user manual, and an empty CD case! When I asked my friend about it, he said he had never opened the box. Someone must have taken the disc before my friend got the package.

This last problem was one I couldn't overcome. I had successfully set up the computer, but obviously couldn't do dictation with no software. In the end I have a surprisingly responsive "virtual" Windows computing solution. Now I just need some reason for using Windows.

Wednesday, January 9, 2008

Hello, Computer?? Apple can you hear me?

There is a classic (among geeks at least) scene in Star Trek IV where the crew of the Enterprise has traveled back in time to the late 20th century. Chief Engineer Scott sits down in front of a Mac computer and says, "Computer. Computer?" Getting no response Dr. McCoy helpfully hands Scott the mouse, which he holds like microphone, "Hello, Computer??" The befuddled 1980's Earthling standing by finally tells him to just use the keyboard ("How quaint"). [Watch on YouTube] While the Mac was not able to respond to voice input, it did bring the mouse-driven graphical user interface revolution to the masses.


Moreover, when Steve Jobs introduced the Macintosh over 20 years ago his demonstration, in part, "let the computer do the talking" via a synthesized voice reading a short speech. This was quite a feat in 1984. However, very little has changed in speech synthesis between then and now. Why has there been so little advancement in voice-based computer interfaces in over two decades? Are the factors finally in place for the next interface revolution to truly put the Personal in PC's. The answer may be yes, and the company poised to lead that change is once again Apple.

The answers to the first question are many. Primarily, voice-based interfaces have stagnated not due to technology constraints but because of a lack of demand. The niche market has been served by software venders like Dragon Systems (now owned by Nuance) who have been able to do voice recognition since the days when the 486 processor ruled. Current iterations showing the feasibility of voice recognition include voice dialing, available on many, even the most inexpensive, cell phones, and the Sync system for making phone calls and controlling digital music players in Ford cars. The lack of demand on desktop systems in the past is largely due to the fact that the majority of computer use took place in the cubicle farms of the American office. Voice interfaces would not fit very effectively into that environment.

As more people spend more time online at home, speech-based interactions make more sense. In addition, many people compose numerous emails, and blog, chat, or Twitter daily. All of these applications would be well served by dictation software. Further, the generation of people having grown up with computers continues to grow. While older people, as a general rule, may be less comfortable with technology, kids and young adults have no aversion to talking to a machine. Perhaps the time is finally right for someone to take this seemingly logical next step in computer interfaces.

If the time is now the company may be Apple. Buoyed by an amazing chain of products since Steve Jobs regained the helm, Apple has shown a repeated ability to take existing technologies and polish and package them in a user-friendly way that brings them to more people. The iPod and iTunes have done it for digital music, OS X for Unix, and now the iPhone/iPod Touch for mobile computing. History has shown Apple to have an interest in improving the user experience. Another major advantage is control of the hardware and software environment and a commitment to open source. Apple has long included built-in microphones on its laptops and all-in-ones. Tweaking these for noise reduction or other speech enhancements would be fairly easy. If they set their engineers to the task, speech could become an intrinsic part of Mac OS.

This is the key to my argument. I don't purport to have done an exhaustive review of the available add-ons that can make a computer voice-activated. Far from it. But that is because this technology should not be an add-on. If I can edit a photo, listen to digital music, browse the internet, and write formatted text using a stock installation of an operating system, I should just as easily be able to search for a file, cue up a song, navigate to a web site, or dictate text without using the keyboard. I'm not saying that the computer should understand complex natural language or that the mouse and keyboard would be replaced entirely. I would be happy to follow a set format for commands and annunciate clearly and separate each word from the others.

Mac OS X even includes some support out of the box for voice recognition and computer speech (my computer tells me the time every half hour). The problem is these features are not highlighted as the way to interact with the computer. Until there is a keynote where Steve Jobs uses spoken commands in a demonstration or there is an Apple ad campaign that shows users talking to their computers, consumers (and therefore developers) won't take speech seriously. But if it were suddenly put forward as part of human interface guidelines a whole new breed of more usable applications could take hold, and the next generation of computer interface could develop. If only Apple is listening.